Consistency first
Start with the invariants your application must protect, especially when an operation fails.
Make failures predictable.
An open-source foundation for explicit failure handling, predictable recovery, and verifiable consistency — designed to grow across programming languages.
An order is submitted.
The external service returns an error.
The outcome is recorded; side effects are not assumed to roll back automatically.
A workflow can span a database, a service, and an external API. A transaction boundary around one component rarely covers them all. Atomic Flow explores what responsible recovery looks like outside that boundary.
Start with the invariants your application must protect, especially when an operation fails.
Make execution boundaries, failures, and recovery outcomes observable instead of implicit.
Validate partial execution and recovery scenarios, not only the happy path.
Keep the execution runtime separate from the tools that challenge it. Build for real failures, not just tidy demonstrations.
PRODUCTION RUNTIME
Define and execute workflows with explicit boundaries and recovery semantics. The public API and supported guarantees are under active design.
@atomic-flow/coreLearn more ↗TESTING & VALIDATION
Explore controlled failure injection and invariant checks. A testing-focused companion — not an extra runtime requirement for production.
@atomic-flow/failure-labLearn more ↗The project aims to document common behavioral expectations, while allowing each language implementation to feel native to its ecosystem.
The first implementation lives in its own repository. Core and Failure Lab are managed together using pnpm workspaces.
Explore the TypeScript repository“The goal isn't to eliminate failures. It's to make their consequences understandable, manageable, and testable.”
Atomic Flow is in its earliest stage. Follow the design, review the evolving documentation, or browse the source. External contributions are temporarily restricted while foundational contracts are established.